How to Model a Fade Period as Returns Normalize

The standard two-stage DCF contains a lie, and it hides in the seam between year five and year six. For five years the model lets a business grow fast and earn a high return on the capital it reinvests. Then, at the stroke of a spreadsheet column, it assumes growth drops to a few percent and the business coasts forever on the same high returns. Real competitive positions don't work like that. They erode on a slope, not off a cliff.

I'll be blunt about where I stand: the usual fix, nudging the terminal growth rate down a point or two, treats the symptom. The terminal growth rate was never the broken part. The broken part is the return on new capital that rides along with it. A company growing 3% while earning 25% on every reinvested dollar is making a claim about the world that almost no business has sustained. Yet that is exactly what a five-year explicit period followed by a Gordon growth terminal value quietly assumes.

What follows is a way to repair the seam with a fade period: a stretch of years in which both growth and incremental returns glide toward their long-run levels. We'll build it with real arithmetic, compare it against the cliff version, and then I'll tell you where I think the method is weaker than it looks. If you want the broader frame first, our guide to calculating the fair value of a stock covers the basics, and our piece on why terminal value is the most fragile part of a DCF explains why this seam matters so much.

Why the Cliff Overstates Intrinsic Value

Start with the identity that links growth to value. Reinvestment is not free, and what it buys depends on the return earned on the new capital. Growth only creates value when that return exceeds the discount rate.

Free cash flow = NOPAT × (1 − g ÷ ROIC), where g is growth and ROIC is the return on new invested capital

Read that carefully. If ROIC equals the discount rate, growth adds nothing: every dollar reinvested earns exactly what investors required, so the business is worth its current NOPAT divided by the discount rate no matter how fast it grows. If ROIC is above the discount rate, growth is valuable. If it's below, growth destroys value. The cliff version of a DCF keeps ROIC high in perpetuity, so the terminal value is paying you for growth that competition would normally have competed away.

This is where the price versus value distinction earns its keep. Price reflects what the market will pay today for a story of durable excess returns. Value is what those returns are worth if they fade at a realistic pace. A model that can't represent fading can't tell the two apart.

Take a large, admired retailer such as Costco (COST), which reported net income of $7,367 million for fiscal 2024, the 52 weeks ended September 1, 2024, per its annual report on Form 10-K. Investors reasonably pay up for a business like that. I haven't rebuilt a Costco model for this post, and I won't pretend the arithmetic below is Costco's. But the question a fade period forces is the right one to ask of any such company: how long, and how fast, do the excess returns decay?

Build the Fade in Three Moves

The method is mechanical. Here are the three moves, in the order I'd do them.

1. Split the forecast into three zones: an explicit period (years 1–5), a fade period (years 6–15), and a terminal period (year 16 onward).
2. In the explicit period, hold growth and ROIC at the levels you can defend from the company's recent filings. In the fade period, move both in a straight line toward their terminal values. Straight lines are crude, but they are easy to audit.
3. Set the terminal ROIC deliberately. The most conservative choice is the discount rate itself, which makes terminal growth worth nothing. A defensible middle ground is a modest premium to it, if you can name the structural reason the premium survives.

A practical way to pick the fade length is to ask what would have to be true for competitors to close the gap. Where the advantage rests on a cost position or a regulatory license, returns can stay elevated for a long time, so a fifteen-year fade might be fair. Where it rests on a product lead in a fast-moving category, five or six years may be generous. You don't need to be right about the number. You need to write it down so the next person can disagree with it, and so that you can see how much your estimate moves when they do.

Every one of those inputs is a hypothesis, so name it. Our guide to choosing a discount rate that actually works covers the first one. The other two, the length of the fade and the terminal ROIC, are the new decisions this method asks you to make explicitly instead of burying them.

A Worked Example With Stated Assumptions

To keep the arithmetic transparent, I'll use an illustrative business, not a real company. Every figure below is an assumption I chose, not a reported number. Start with year-0 NOPAT of $100 million and a 9% discount rate. In years 1–5, growth is 8% a year and ROIC on new capital is 25%, so the reinvestment rate is 8% ÷ 25% = 32% of NOPAT. Terminal growth is 3%.

Now run four versions, with all values in millions of dollars. Under the cliff with ROIC held at 25% forever, the five explicit years contribute about $331 million and the terminal value contributes about $1,443 million in today's money, for a total near $1,773 million. Terminal value is roughly 81% of the answer. Under the cliff with ROIC dropping to 9% at year six (growth is worthless afterward), the total is about $1,424 million.

Now add the fade. Keep the same five years, then glide growth from 8% to 3% and ROIC from 25% to 9% across years 6–15. The explicit and fade years together contribute about $887 million, the terminal value about $769 million in today's money, and the total lands near $1,656 million. If the terminal ROIC settles at 12% instead of 9%, the total is about $1,779 million.

Illustrative value ($ millions): cliff at 25% ROIC ≈ 1,773; fade to 12% ≈ 1,779; fade to 9% ≈ 1,656; cliff at 9% ≈ 1,424

Two things stand out. First, the spread between the most generous and most conservative treatments is about 25%, from a single assumption about how returns behave after year five. That's bigger than most debates over a one-point change in the discount rate. Second, the fade moves value out of the terminal bucket and into the years you can actually observe: terminal value falls from roughly 81% of the total to about 46% in the fade-to-9% case. That's a more honest distribution of where the value comes from.

Notice also the coincidence at the top of the table. The fade-to-12% case lands almost exactly on the cliff-at-25% number. That isn't a law of nature. It's two offsetting errors in one direction, and I'd distrust any model that appears to confirm itself that neatly.

It also helps to run the example backward as a sanity check. If the cliff-at-25% value of $1,773 million is what the market is paying, then the price embeds the belief that the business earns a 25% return on new capital in perpetuity. Put that way, most investors would balk. The same price is far easier to defend if you believe the fade ends near 12%, and that's a conversation about evidence, such as segment returns, reinvestment history and competitor behavior in the filings, instead of a conversation about a single magic terminal multiple.

Where the Fade Method Is Weaker Than It Looks

I'm less sure about this method than the tidy numbers suggest. A linear fade is a convenience, not a finding. Real return decay is lumpy: a patent cliff, a regulatory change or a new entrant can compress returns in one year, while a network effect can hold them for decades. The data on how long excess returns actually persist across industries is mixed, and I'd be wary of anyone who quotes a single fade length as settled fact.

The method also gives you more dials to turn, and more dials means more room to reverse-engineer a conclusion. If you catch yourself lengthening the fade until the model matches the current share price, stop. You've stopped valuing the business and started rationalizing the price. A cleaner use is the opposite direction: pair this with a reverse DCF that backs out the growth rate the market is pricing in, then ask how long the fade must be to justify it.

And a caution on scope. Judging whether a company's advantage will last is a qualitative moat question, and it deserves its own analysis. The fade period only translates that judgment into numbers. It doesn't replace it. Treat the output as a compass, not a GPS coordinate: it tells you which direction the evidence points and roughly how far you'd be from a margin of safety, not an address.

Key Takeaways

• A cliff between the explicit period and the terminal value assumes high returns on new capital last forever. A fade period models the erosion that competition usually brings.
• Growth creates value only when ROIC exceeds the discount rate. Setting terminal ROIC equal to the discount rate is the most conservative choice and a useful floor.
• In the illustrative example, one assumption about post-year-five returns moved value by about 25%, and a fade cut terminal value's share of the total from roughly 81% to 46%.
• A fade is a hypothesis, not a measurement. Name the length and the terminal ROIC, test them against a reverse DCF, and don't lengthen the fade just to match the market price.